The community functional trait characterization in species-rich ecosystems faces sampling trade-offs on species abundances representativeness and trait variability. Overall, sampling designs for dominant species have received broad attention by using community-weighted mean (CWM) to indicate locally optimal phenotypes. CWM, as a descriptor of optimal phenotypes, does not consider trait variability, failing in to explain the mechanisms behind of community trait composition. In particular, when viable co-existing strategies promote divergence, or at fine scales where ecosystems have similar filtering pulses, as is expected in Tropical Dry Forests. We compared different trait sampling designs, vary the species abundance representativeness and trait variability, to evaluate differences in the level of trait community characterization and the trait-environment and trait-biomass relationships in TDF. Following and abundance-weighted trait sampling design, we intensively sampled 15 functional traits in 321 species of ten 1-ha permanent plots in TDF. We sampled a least one individual per species per plot (in the case of ‘rarer’ species) up to 12 individuals when abundant (N=1391 tree individuals). We also monitored stem biomass growth for 19,740 trees for the sampled species, between 2013 and 2017. To evaluate the abundance representativeness, we ran linear correlations between the total species abundances per community and those for the abundance-weighted trait sampling and dominant ones. To assess the main sources of trait variability, we performed a nested variance partitioning with individuals, populations, species, and communities as the ecological scales. We performed linear and linear mixed-effects models for the sampling designs to test the trait-environment and trait-biomass relationships. We found that (i) sampling designs considering only dominant species did not adequately reflect the species abundance representativeness in TDF. (ii) Differences within communities explained a higher proportion of total trait variability than among them. (iii) Trait-environment and trait-biomass had consistently stronger relationships when samplings improve the species abundances representativeness and trait variability. Our results indicate that abundance-weighted trait sampling designs may be useful to reconciling the trade-offs between species abundances representativeness and trait variability in samplings characterizing community trait composition, and to detect trait-environment and trait-biomass relationships in TDF.